Recent Acceleration of Denman Glacier (1972-2017), East Antarctica, Driven by 2 Grounding Line Retreat and Changes in Ice Tongue Configuration
Total Page:16
File Type:pdf, Size:1020Kb
https://doi.org/10.5194/tc-2020-162 Preprint. Discussion started: 6 July 2020 c Author(s) 2020. CC BY 4.0 License. 1 Recent acceleration of Denman Glacier (1972-2017), East Antarctica, driven by 2 grounding line retreat and changes in ice tongue configuration. 3 4 Bertie W.J. Miles1*, Jim R. Jordan2, Chris R. Stokes1, Stewart S.R. Jamieson1, G. Hilmar 5 Gudmundsson2, Adrian Jenkins2 6 1Department of Geography, Durham University, Durham, DH1 3LE, UK 7 2Department of Geography and Environmental Sciences, Northumbria University, Newcastle upon Tyne, NE1 8 8ST, UK 9 *Correspondence to: [email protected] 10 11 Abstract: Denman Glacier is one of the largest in East Antarctica, with a catchment that 12 contains an ice volume equivalent to 1.5 m of global sea-level and which sits in the Aurora 13 Subglacial Basin (ASB). Geological evidence of this basin’s sensitivity to past warm periods, 14 combined with recent observations showing that Denman’s ice speed is accelerating, and its 15 grounding line is retreating along a retrograde slope, have raised the prospect that it could 16 contribute to near-future sea-level rise. In this study, we produce the first long-term (~ 50 years) 17 record of past glacier behaviour (ice flow speed, ice tongue structure, and calving) and combine 18 these observations with numerical modelling to explore the likely drivers of its recent change. 19 We find a spatially widespread acceleration of the Denman system since the 1970s across both 20 its grounded (17 ±4% acceleration; 1972-2017) and floating portions (36 ±5% acceleration; 21 1972-2017). Our numerical modelling experiments show that a combination of grounding line 22 retreat, ice tongue thinning and the unpinning of Denman’s ice tongue from a pinning point 23 following its last major calving event are required to simulate an acceleration comparable with 24 observations. Given its bed topography and the geological evidence that Denman Glacier has 25 retreated substantially in the past, its recent grounding line retreat and ice flow acceleration 26 suggest that it could be poised to make a significant contribution to sea level over the coming 27 century. 28 29 1. Introduction 30 Over the past two decades, outlet glaciers along the coastline of Wilkes Land, East Antarctica, 31 have been thinning (Pritchard et al., 2009; Flament and Remy, 2012; Helm et al., 2014; 1 https://doi.org/10.5194/tc-2020-162 Preprint. Discussion started: 6 July 2020 c Author(s) 2020. CC BY 4.0 License. 32 Schröder et al., 2018), losing mass (King et al., 2012; Gardner et al., 2018; Shen et al., 2018; 33 Rignot et al., 2019) and retreating (Miles et al., 2013; Miles et al., 2016). This has raised 34 concerns about the future stability of some major outlet glaciers that primarily drain the Aurora 35 Subglacial Basin (ASB), particularly Totten, Denman, Moscow University and Vanderford 36 Glaciers (Fig. 1). This is because their present day grounding lines are close to deep retrograde 37 slopes (Morlighem et al., 2020), meaning there is clear potential for marine ice sheet instability 38 and future rapid mass loss (Weertman, 1974; Schoof, 2007), unless ice shelves provide a 39 sufficient buttressing effect (Gudmundsson, 2013). Geological evidence suggests that there 40 may have been substantial retreat of the ice margin in the ASB during the warm interglacials 41 of the Pliocene (Williams et al., 2010; Young et al., 2011; Aitken et al., 2016; Scherer et al., 42 2016), which potentially resulted in global mean sea level contributions of up to 2 m from the 43 ASB (Aitken et al., 2016). This is important because these warm periods of the Pliocene may 44 represent our best analogue for climate by the middle of this century under unmitigated 45 emission trajectories (Burke et al., 2018). Indeed, numerical models now predict future sea 46 level contributions from the outlet glaciers which drain the ASB over the coming decades to 47 centuries (Golledge et al., 2015; Ritz et al., 2015; DeConto and Pollard, 2016), but large 48 uncertainties exist over the magnitude and rates of any future sea level contributions. 49 At present, most studies in Wilkes Land have focused on Totten Glacier which is losing mass 50 (Li et al., 2016; Mohajerani et al., 2019) in association with grounding line retreat (Li et al., 51 2015). This has been attributed to wind-forced warm Modified Circumpolar Deep Water 52 accessing the cavity below Totten Ice Shelf (Greenbaum et al., 2015; Rintoul et al., 2016; 53 Greene et al., 2017). However, given our most recent understanding of bedrock topography in 54 Wilkes Land, Denman Glacier provides the most direct pathway to the deep interior of the ASB 55 (Gasson et al., 2015; Brancato et al., 2020; Morlighem et al., 2020). Moreover, a recent mass 56 balance estimate (Rignot et al., 2019) has shown that, between 1979 and 2017, Denman 57 Glacier’s catchment may have lost an amount of ice (190 Gt) broadly comparable with Totten 58 Glacier (236 Gt). There have also been several reports of inland thinning of Denman’s fast- 59 flowing trunk (Flament and Remy, 2012; Helm et al., 2014; Young et al., 2015; Schröder et 60 al., 2018) and its grounding line has retreated over the past 20 years (Brancato et al., 2020). 61 However, unlike Totten and other large glaciers which drain marine basins in Antarctica, there 62 has been no detailed study analysing any changes in its calving cycle, velocity or ice tongue 63 structure. This study reports on a range of remote sensing observations from 1962 to 2018 and 64 then brings these observations together with numerical modelling to explore the possible 2 https://doi.org/10.5194/tc-2020-162 Preprint. Discussion started: 6 July 2020 c Author(s) 2020. CC BY 4.0 License. 65 drivers of Denman's long-term behaviour. The following section outlines the methods (section 66 2) used to generate the remote sensing observations (section 3) and we then outline the 67 numerical modelling experiments (section 4) that were motivated by these observations, 68 followed by the discussion (section 5). 69 70 2. Methods 71 2.1 Ice front and calving cycle reconstruction 72 We use a combination of imagery from the ARGON (1962), Landsat-1 (1972-74), Landsat 4- 73 5 (1989-1991), RADARSAT (1997) and Landsat 7-8 (2000-2018) satellites to create a time 74 series of ice-front position change from 1962-2018. Suitable cloud-free Landsat imagery was 75 first selected using the Google Earth Engine Digitisation Tool (Lea, 2018). Changes in ice- 76 front position were calculated using the box method, which uses an open ended polygon to take 77 into account any uneven changes along the ice-front (Moon and Joughin, 2008). To supplement 78 the large gap in the satellite archive between 1974 and 1989 we use the RESURS KATE-200 79 space-acquired photography from September 1984. This imagery is hosted by the Australian 80 Antarctic Data Centre, and whilst we could not access the full resolution image, the preview 81 image was sufficient to determine the approximate location of the ice-front and confirm that a 82 major calving event took place shortly before the image was acquired (Fig. S1). 83 84 2.2 Velocity 85 Maps of glacier velocity between 1972 and 2002 were created using the COSI-Corr (CO- 86 registration of Optically Sensed Images and Correlation) feature-tracking software (Leprince 87 et al., 2007; Scherler et al., 2008). This requires pairs of cloud-free images where surface 88 features can be identified in both images. We found three suitable image pairs from the older 89 satellite data: Nov 1972 – Feb 1974, Feb 1989 – Nov 1989, and Nov 2001 – Dec 2002. We 90 used a window size of 128 x 128 pixels, before projecting velocities onto a WGS 84 grid at a 91 pixel spacing of 1 km. 92 To reduce noise, we removed all pixels where ice speed was greater than ±50% the MEaSUREs 93 ice velocity product (Rignot et al., 2011b), and all pixels where velocity was <250 m yr-1. Errors 94 are estimated as the sum of the co-registration error (estimated at 1 pixel) and the error in 3 https://doi.org/10.5194/tc-2020-162 Preprint. Discussion started: 6 July 2020 c Author(s) 2020. CC BY 4.0 License. 95 surface displacement (estimated at 0.5 pixels). This resulted in total errors ranging from 20 to 96 73 m yr-1. Annual estimates of ice speed between 2005-2006 and 2016-2017 were taken from 97 the annual MEaSUREs mosaics (Mouginot et al., 2017). These products are available at a 1 98 km spatial resolution and are created from the stacking of multiple velocity fields from a variety 99 of sensors between July and June in the following year. To produce the ice speed time-series, 100 we extracted the mean value of all pixels within a defined box 10 km behind Denman’s 101 grounding line (see Fig. 3). To eliminate any potential bias from missing pixels, we placed 102 boxes in locations where all pixels were present at each time step. 103 We also estimated changes in the rate of ice-front advance between 1962 and 2018. This is 104 possible because inspection of the imagery reveals that there has been only one major calving 105 event at Denman during this time period because the shape of its ice front remained largely 106 unchanged throughout the observation period. Similar methods have been used elsewhere on 107 ice shelves which have stable ice fronts e.g.